The UPC bar code marked its 50th birthday earlier this year. GS1 is the not-for-profit enterprise that maintains that global standard for product identification, and GS1 US is one of 116 country-based GS1 member organizations.
GS1 barcodes are scanned over 10 billion times a day — and those scans go far beyond your Kroger or Publix grocery checkout. GS1 standards organize the supply chains of sectors like apparel, food service, and healthcare. In a nutshell, this is bedrock capability for the global supply chain. But even though the numbering standard itself has been mostly unchanged since its first scan in 1974, nearly everything around barcodes has been innovating and changing.
Earlier this year, in Orlando, I had the opportunity to sit down with Bob Czechowicz, Senior Director of Innovation at GS1 US, who was in town for the annual GS1 Connect event. The summit brought together ecosystem members from the hardware, software, retail, and government communities to discuss industry innovations like digital watermarks and 2D barcodes. Czechowicz shared his team’s journey in innovating the future of AI as part of our latest research initiative, The Future of the Innovation Team.
• • •
What’s the make-up of your current innovation team?
At our headquarters in Ewing, New Jersey, GS1 US has a headcount of about 200 people. I lead an innovation team [that] collectively has a responsibility for running a variety of pilots. Our goal with these pilots is to locate some promising intersection of technology and market. We’re looking for ways that GS1 standards can improve supply chain business processes, or support regulatory requirements…
We’re essentially the chief hypothesis hatchery. We run about 10-15 experiments per year. It’s our job to generate those “how might we” statements and coax potential testable solution ideas from them. Bottom line, GS1 US looks to our group to continually scan for and evaluate many emerging technologies that can impact the supply chain in positive ways.
What are some boundary conditions around the type of projects you take on, or say no to?
As a not-for-profit and a standard bearer for the industry, we’re charged with demonstrating the art of the possible in a way that doesn’t favor one startup or one company but rather in a way that “lifts all boats”; i.e., that makes a new set of solutions accessible to all players in the ecosystem. We’ve hosted a startup pitch competition at GS1 Connect events. When you consider the various industry stakeholders we work with — manufacturers, wholesalers/distributors, retailers, restaurants, healthcare providers, government agencies, trade associations, solution providers and technology startups—we support events and cooperative engagements that network all these folks together.
How would you characterize your own GS1 US approach to innovation?
We have a defined process that we follow that borrows from various playbooks. Frankly, I often discover through events like InnoLead webinars that we’re not alone, thank goodness! We are all doing similar things and learning as we go. I could say that our approach is very influenced by Eric Ries lean startup style innovation; i.e., early and frequent testing, and so on. We know that identifying juicy problems is the hard part. When we do it right, then we’re not going around with a hammer in our hand in search of nails.
GenAI may be the biggest hammer that any of us have ever held in our hands. How is your team trying not to see everything as AI nails?
Our firm is responsible for maintaining one of the world’s most important global commerce standards so first thing I’d say is we’re not looking to rush into putting AI out into the field at the moment. Hallucination on a standard could cause some issues! We’re more in a mode of de-risking, discovery and exploration, especially around internal use cases. So right off the bat, again, we’re looking for good problems to solve, not where we can just apply GenAI.
What have you put in motion internally thus far?
Towards the end of 2023 we were chartered as an “AI Tiger Team.” Myself and our VP of Technology, Bill Fraser, were given the task of co-leading a team to identify implementable internal use cases — things that we could put in front of our internal customers. We held a three-hour session for 60 people using Miro boards, and even found we were able to leverage some of the AI capabilities of Miro in the process, like their auto affinity clustering and idea summarization. After a few rounds of those, perhaps unsurprisingly, most roads led to chatbots. We also looked at fine-tuning our own models, supporting them in reducing hallucinations by implementing retrieval augmented generation (RAG).
What are the next steps on your Tiger Team’s AI journey?
As a firm, we have onboarded everyone to Microsoft Copilot. We’re all-in on folks using the capabilities in Office365. For example, one cool thing I use often is a simple prompt like, “Tell me all my actions this week from my emails.” And it works. Events can then auto- magically appear on my calendar.
The biggest concern of any enterprise [with AI] is, does it leak? As in, will our confidential information wind up in someone’s environment, or will our information be used to train other models?
We are also presently establishing clear company guidelines and guardrails. The biggest concern of any enterprise is, does it leak? As in, will our confidential information wind up in someone’s environment, or will our information be used to train other models? That’s a baseline for us. Also, we want to adhere to our record retention policies. Like, you close the browser window, and the data is gone. Our legal team is quite actively involved in these discussions.
As for what’s next, we see AI as an opportunity to evangelize on data quality and data hygiene. If you’re not doing data well, you will not be ready to embrace this transformation. If your data house is in order, in principle, you can build cleaner and more purposeful models.
And even though we’re focused on internal use cases as I said before, if we get our data clean and structured, as we move forward, we certainly can envision pilots around things like predicting the next consumer purchase based on prior data. Or, for that matter, we’re already doing some investigation into using AI to predict product classifications. This is a challenging task, as there are often many different ways to classify a product. But AI can help us to identify the most likely classification, which can save our members time and money. In other words, AI could infer that this Global Product Classification (GPC) should be tagged “beverage” based on historical data.
What do you count as your AI Tiger Team’s key wins thus far?
AI policy [and] drafting good recommendations might be our biggest win. We took that on squarely and made some key presentations to leadership. We were transparent about the positive potentials and the risks and hazards, like, you still need to check your own work, and not take AI outputs as endpoints.
I’d say we are doubling down on storytelling. [At the] end of the day, innovation is communication to a variety of leaders.
Final question, beyond AI-specific things, are there new outcomes you’re being expected to deliver, or new KPIs you’re being measured on?
Yes. Historically we were measured and evaluated solely based on the number of pilots we ran each year, which as I said before can vary but it’s roughly 10-15 each year. Now we are reporting not just the number, but for each pilot providing an accompanying strategy document, that centers [on] the learnings we gained, including clear stories around the stop/kills. All of that folds into our end of year strategy narrative. And finally, I’d say we are doubling down on storytelling. [At the] end of the day, innovation is communication to a variety of leaders. If we can’t tell a good compelling story, ideas with great potential can be left on the table; so great visual and crisp communication is a priority for us as well.
